Utilizing the similarity preserving properties of self-organizing maps in vector quantization of images

The self-organizing map (SOM) algorithm creates a topologically ordered mapping from the input space to map nodes. The mapping has the special property that the neighborhood relations between the input samples are preserved to the output space. In this paper it is shown that the similarity preserving property of the SOM can be used advantageously in image vector quantization applications, either to increase the error tolerance for transmission errors, or to increase the compression efficiency